UNI-MB - logo
UMNIK - logo
 
(UM)
  • Inductive learning of relations from noisy examples
    Lavrač, Nada ; Džeroski, Sašo
    The success of inductive learning algorithms in practical domains largely depends upon their ability to handle imperfect data. Most of the successful systems use a propositional attribute-value ... language to represent training examples and induced concept descriptions. On the other hand, systems which learn relational descriptions in the form of logic programs have only recently addressed the problem of learning from noisy data. LINUS and FOIL are two such systems which are based on approaches known from attribute-value learning algorithms. The paper gives a comparison of the two systems, as well as an empirical comparison of their performance on the problem of learning illegal chess endgame positions, both from non-noisy examples and examples that have been corrupted with noise.
    Vir: Inductive logic programming (Str. 495-516)
    Vrsta gradiva - članek, sestavni del
    Leto - 1992
    Jezik - angleški
    COBISS.SI-ID - 2715670

Vnos na polico